PARS (Parallel Systems) Research Group, Izmir Institute of Technology (IZTECH)
|Research Assistant
Izmir, , Türkiye
Summary
Served as a Research Assistant within the Parallel Systems Research Group, focusing on GPU performance optimization, reproducible research environments, and integrating deep learning tools for heterogeneous CPU-GPU systems.
Highlights
Conducted CUDA-based performance exploration and benchmarking on desktop-class discrete GPUs and embedded Jetson Nano devices, identifying key optimization opportunities.
Profiled kernels using Nsight Systems and Nsight Compute to analyze memory access patterns, occupancy, and execution bottlenecks, leading to targeted performance improvements.
Built Dockerized GPU environments to ensure reproducible experiments and remote access, significantly enhancing support for student and faculty research within the PARS group.
Integrated NVIDIA Deep Learning Institute (DLI) teaching kits into benchmarking and educational pipelines, streamlining the development of heterogeneous CPU-GPU systems.
Prepared an internal technical report on memory-optimized GPU scheduling and performance modeling, contributing to advanced research documentation and knowledge dissemination.